From owner-freebsd-hackers  Tue Aug 18 13:37:46 1998
Return-Path: <owner-freebsd-hackers@FreeBSD.ORG>
Received: (from majordom@localhost)
          by hub.freebsd.org (8.8.8/8.8.8) id NAA21995
          for freebsd-hackers-outgoing; Tue, 18 Aug 1998 13:37:46 -0700 (PDT)
          (envelope-from owner-freebsd-hackers@FreeBSD.ORG)
Received: from mailhub.scl.ameslab.gov (mailhub.scl.ameslab.gov [147.155.137.127])
          by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id NAA21990
          for <Hackers@freebsd.org>; Tue, 18 Aug 1998 13:37:44 -0700 (PDT)
          (envelope-from ghelmer@scl.ameslab.gov)
Received: from demios.ether.scl.ameslab.gov ([147.155.137.54] helo=demios.scl.ameslab.gov)
	by mailhub.scl.ameslab.gov with smtp (Exim 1.90 #1)
	id 0z8sQs-0006lF-00; Tue, 18 Aug 1998 15:32:42 -0500
Date: Tue, 18 Aug 1998 15:36:57 -0500 (CDT)
From: Guy Helmer <ghelmer@scl.ameslab.gov>
To: "Alton, Matthew" <Matthew.Alton@anheuser-busch.com>
cc: "'Hackers@FreeBSD.ORG'" <Hackers@FreeBSD.ORG>
Subject: Re: DIY Supercomputers
In-Reply-To: <31B3F0BF1C40D11192A700805FD48BF90177663A@STLABCEXG011>
Message-ID: <Pine.SGI.3.96.980818145941.20758D-100000@demios.scl.ameslab.gov>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-freebsd-hackers@FreeBSD.ORG
Precedence: bulk
X-Loop: FreeBSD.ORG

On Tue, 18 Aug 1998, Alton, Matthew wrote:

> Has anybody taken a look at the possibility of porting Beowulf
> http://www.beowulf.org
> to FreeBSD?  Or are we clustering like the very wind our own selves someplace?

The basic message-passing software on which parallel processing
applications are typically built, MPI (e.g. MPICH or LAM implementations) 
or PVM, ought to work on FreeBSD (I've personally used MPICH on FreeBSD). 
I haven't personally needed for some of the Beowulf enhancements (Ethernet
"channel bonding", vm pre-pager), although the unified PID space would be
useful, and FreeBSD has kernel support for reading the Pentium performance
counters. 

The biggest objection I've heard to using FreeBSD for parallel processing
clusters was the lack of a FreeBSD version of a certain commercial Fortran
compiler.  FreeBSD (with net.inet.tcp.delack_enabled=0 in 3.0-current) 
ought to have better network performance over the entire spectrum of
message sizes than Linux 2.0.x, so I want to see whether this has an
observable effect on parallel apps.

Guy Helmer, Graduate Student, Iowa State University Dept. of Computer Science 
Research Assistant, Ames Laboratory       ---         ghelmer@scl.ameslab.gov
http://www.cs.iastate.edu/~ghelmer


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message